Knowledge Graph Fact Linking for Text Ambiguity Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to accurately link and disambiguate entities and relations in textual data due to ambiguity across different documents, leading to missed crucial information.
Innovation Solution
A method that ranks and links textual facts to a reference knowledge graph, determining whether each element of the triple data structure should be linked to an existing entry or become a new entry, using pre-trained language models and graph-based embeddings to generate an enriched knowledge graph.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If textual facts are presented as triples without linking to knowledge graphs, then the format is simple and easy to process, but ambiguity in entities and relations cannot be resolved leading to missed information
Solution Approach 1:
The patent introduces a knowledge graph as an intermediary structure that mediates between textual facts and their meanings. The knowledge graph serves as a reference framework that disambiguates entities and relations by providing standardized identifiers and relationships, thereby resolving the contradiction between simple triple format and accurate information representation.
Solution Approach 2:
The patent performs preliminary actions by pre-processing textual data to extract triples and then systematically linking them to the knowledge graph before final output. This preliminary linking process ensures that ambiguity is resolved in advance, improving information accuracy without complicating the final presentation format.
2Loss of information
If entities and relations are linked to knowledge graphs, then ambiguity is reduced and information completeness improves, but the complexity of processing and system architecture increases
Solution Approach 1:
The knowledge graph serves multiple functions simultaneously: it acts as a reference for entity disambiguation, provides relationship context, enables information completeness checking, and facilitates cross-document linking. This multi-functionality reduces the need for separate processing systems, thereby managing complexity while improving information completeness.
Solution Approach 2:
The patent creates a simplified representation of the knowledge graph structure that can be copied and applied across different textual datasets. The knowledge graph schema and linking mechanisms can be replicated for different domains, reducing system complexity through standardization while maintaining information completeness across various applications.
3Measurement precision
If all textual facts are manually verified and linked, then accuracy is maximized, but processing time and efficiency decrease
Solution Approach 1:
The system performs self-service by automatically extracting triples from text and autonomously linking them to the knowledge graph without requiring manual verification for each link. The automated linking process maintains high accuracy through algorithmic matching while dramatically improving processing efficiency compared to manual verification methods.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously refines its linking accuracy by comparing extracted triples against the knowledge graph structure and adjusting its matching algorithms. This feedback loop enables the system to maintain high measurement precision in linking accuracy while processing large volumes of text efficiently through iterative improvement.
Data Source
AI summary
A method for disambiguating textual facts by linking the textual facts to a reference knowledge graph and introducing entities and relations for the textual facts includes ranking the textual facts from textual data. Each of the textual facts is in a triple data structure comprising a subject, a relation, and an object. The method further includes determining whether each element of the triple data structure should be linked to an existing entry within the reference knowledge graph or become a new entry within the data element, and generating an enriched knowledge graph based on the determination. A ranking module using machine learning can be used. The method can help users for decision making and can be used in a variety of applications including, but not limited to, several use cases in drug development, material synthesis, inventory of items, detection of suspects from textual messages, and medical/healthcare.